Hong Kong – Leon Lei, who teaches information science within the school of schooling at The College of Hong Kong (HKU), not too long ago produced a textbook – in 30 hours.
Utilizing a mixture of generative AI and different instruments, Lei turned transcripts and slides from a collection of on-line lessons he taught throughout pandemic lockdowns into textual content (15 hours), then edited and compiled it into a ten,000-word course e-book (one other 15 hours). He additionally transformed chapters to thoughts maps – diagrams which present ideas in visible type – and created video clips.
“College students have numerous studying types,” stated Lei, who’s working AI clinics throughout the college on utilizing AI instruments for instructing. “Some need to hear, watch. Some need a thoughts map first. Earlier than this, I didn’t have time to discover.”
Generative AI is inflicting lecturers to rethink how they train and the way they will put together college students for the longer term. Directors are reframing what universities must be instructing that future employers will need.
“When did we final have this type of shake-up?” stated Pauline Chiu, affiliate vice chairman for instructing and studying at HKU. “We’ve obtained the eye of lecturers, dad and mom, college students. It’s a chance to reinvent our instructing. Now that’s a giant optimistic.”
Generative AI instruments – constructed on giant language fashions (LLMs) that synthesize huge troves of information to generate textual content, code, photos and extra – are seen as the most important technological leap because the internet browser and sensible telephones. However whereas the know-how is highly effective, it may possibly ship imperfect outcomes and studying establishments have been grappling in the previous few months over easy methods to deploy it responsibly – if in any respect.
The arrival of generative AI is elevating greater questions on what kind of future universities must be getting ready college students for. “What ought to we be instructing in college alongside it? What do future workers want? What sort of different human abilities will we have to be instructing? How do you collaborate with different human beings, what about relationship constructing?” stated Chiu.
HKU, a research-led complete college based in 1911, is the oldest college in Hong Kong and identified for its medical faculty. It’s ranked as the highest college in Hong Kong and 35th globally, in keeping with the Occasions Greater Schooling World College Rankings 2024.
HKU initially instituted a ban earlier within the yr on utilizing generative AI instruments. “We knew it was momentary,” stated Chiu. In February, a activity pressure made up of employees, college students and technologists started assembly weekly to debate the implications of the brand new know-how.
When Microsoft Azure OpenAI Service, powered by OpenAI’s GPT, grew to become typically out there in January, the college’s IT division acted first. “We stated AI is the longer term,” stated Flora Ng, chief data officer and college librarian. “It’s what we have to pursue to boost analysis, instructing and studying.”
HKU was already utilizing Microsoft’s options and the IT division made an uncommon choice to go forward and fund Azure OpenAI Service to employees solely from April to June, so they might check it out and perceive the affect of generative AI. Often, new IT funding includes tender paperwork with detailed necessities and buy-in from numerous departments – which may take months.
On this case, “no person knew what the overall necessities must be,” stated Ng. “Our IT division stated we’ll take a threat; we’re going to fund it. The technique for me was to shortly undertake, then if it fails, to pivot.”
In June, the HKU senate formally endorsed a generative AI coverage that established it as a “fifth literacy” for college students, alongside oral, written, visible and digital literacy. On the finish of August, HKU and 7 different universities in Hong Kong introduced they have been making Azure OpenAI Service out there to all employees and college students, with few restrictions, when the brand new faculty semester began in September.
“Our stance is to embrace it,” stated Chiu, including that it’s as much as lecturers in the event that they wished to restrict use for his or her programs or sure assignments.
“There could also be conditions by which we would like college students to study the fundamentals by limiting the usage of generative AI,” stated Chiu. “However it is going to be as much as the lecturers to make that call.”
The college has rolled out a number of generative AI chatbots, constructed with Azure OpenAI Service. An IT helpdesk chatbot solutions easy queries, liberating employees as much as take care of extra sophisticated points. One other chatbot offers with administrative questions, similar to how to join a course, and one other one on undergraduate course choice.
Employees and college students may also entry a extra normal HKU chatbot for instructing and studying.
Early utilization statistics have been encouraging. Within the first 20 days, because it was launched on September 1, greater than 10 p.c of the coed inhabitants of 36,400 have used the final HKU chatbot. About 17 p.c of the employees inhabitants of 13,100 have executed the identical. The IT helpdesk chatbot in flip acquired 1,276 inquiries between August 21 and September 13.
To guard person information and privateness, the chatbots don’t maintain any information on queries. “We don’t take a look at what they ask. We don’t maintain any information,” stated Ng.
College students are taught primary generative AI literacy in AI workshops run by HKU’s Educating and Studying Innovation Centre. They study that outcomes should not at all times excellent, however that the HKU chatbot might be a good suggestion generator. Neither is it a search engine; it’s a language synthesizer. They’re instructed to at all times verify authentic sources for accuracy. And so forth.
The massive concern, after all, is that college students develop into too reliant on generative AI to finish assignments with out actually understanding the supplies. Nonetheless, HKU lecturers won’t be counting on AI detection instruments as a result of they aren’t correct or dependable at present. There’s a chance of false negatives and false positives, which might result in a pupil being wrongly accused of dishonest.
As a substitute, lecturers are requested to reinvent evaluation, stated Cecilia Chan, director of HKU’s Educating and Studying Innovation Centre. “Precisely what do we would like the coed to study?” stated Chan. “Take into consideration the training course of, outcomes and expertise, that’s what is necessary.”
A instructor might, for instance, ask for a abilities demonstration or an oral presentation as an alternative of an essay. Or they might ask the chatbot to generate various essays and ask college students to critique them. Are there factual errors? College students might add their opinions and perhaps generate an essay plan. It’s a kind of reverse engineering of an essay, the place “you may nonetheless have all the training targets of an essay,” stated Chan.
A pupil may very well be requested to exhibit competency by hands-on work at completely different stations like within the medical faculty’s Goal Structured Medical Examination, stuff that AI can not do.
Chan stated she herself makes use of generative AI instruments “like a private assistant,” together with to reply the numerous emails she will get asking her for interviews and to talk at conferences. To those that fear about counting on it an excessive amount of, she gives a comparability.
“Are you able to think about life with out considered one of these?” she asks, waving her smartphone. “That’s what we’re getting used to.”
College students are already figuring all this out for themselves.
Lai Yan Ying, also called Cheri, is a fourth-year pupil majoring in linguistics. She stated she wouldn’t use it for writing an essay however thinks it’s truthful to make use of it to generate concepts, similar to questions for latest analysis challenge the place she interviewed somebody about their expertise studying English.
“I don’t suppose we will simply use ChatGPT for every little thing,” stated Lai. “Typically, I simply favor to go to a library and seize a e-book.”
For Yan Wing Lam, a fourth-year engineering main, generative AI is much less of a thoughts shift, “In engineering, we’re fairly into AI already. It’s similar to a instrument to me.”
When Lai and Yan each labored on a latest challenge collectively, they encountered each the promise and the restrictions of the instrument. For a course referred to as Digitizing Cultural Heritage in Larger China, they determined to make use of OpenAI’s DALL·E 3 picture generator by way of Azure OpenAI Service on the HKU chatbot to create photos of Chinese language legendary creatures – with their teacher’s blessings.
They have been solely partially profitable.
A immediate for “blue dragon with horns and claws” efficiently introduced forth an image of Qing Lengthy, an azure dragon god in Chinese language mythology. Nonetheless, it took just a few tries to generate a usable image of Chi Ru, a fish with a human face.
Makes an attempt to conjure up a likeness of Xiang Liu, a nine-headed monster snake, saved returning photos of single-headed snakes. Lai ended up making her personal digital drawing of Xiang Liu, which took about half an hour, versus simply seconds utilizing DALL·E 3.
The four-person crew obtained an A on the challenge.
Prime picture: Leon Lei, who teaches information science within the school of schooling, is utilizing generative AI instruments to create thoughts maps and quick movies for college students with completely different studying types. Picture by Lam Hei Chun for Microsoft.